It was the Task.Run implementation.
One thing I noticed was that the Task methods were getting top performance with GH_Structure and without this error, but with Lists I had to manually subdivide the iterations in 8 tasks (1 per each CPU thread) to get 100% CPU usage.
I have already deleted that part of the code and replaced with Parallel looping but in the version that had the best performance (Task.Factory did not produce better performance), it looked something like this:
List<Mesh[]> meshList = new List<Mesh[]>();
//Queue up tasks ----------------------------------------------
Task<List<Mesh[]>> t0 = Task.Run(() => DrawGeometryT0());
Task<List<Mesh[]>> t1 = Task.Run(() => DrawGeometryT1());
...
//Halt until all geometry has been created -------------------
List<Task> taskList = new List<Task>{t0, t1, ... };
Task.WaitAll(taskList.ToArray());
//Results assembly --------------------------------------------
foreach (Mesh[] mshA in t0.Result) meshList.Add(mshA);
foreach (Mesh[] mshA in t1.Result) meshList.Add(mshA);
...
//Define methods ----------------------------------------------
List<Mesh[]> DrawGeometryT0() {
for (iteration logic - segment1) {
Mesh geometry creation;
}
return meshArrayList0;
}
List<Mesh[]> DrawGeometryT1() {
for (iteration logic - segment2) {
Mesh geometry creation;
}
return meshArrayList1;
}
